Fusion positioning noise model optimization method and device, equipment and readable storage medium

Abstract

The application provides a fusion positioning noise model optimization method and device, equipment and a readable storage medium, which is based on an approximate Gaussian model, can optimize GNSS noise model in real time in a dynamic environment, evaluates speed error according to the optimized GNSS real-time measurement noise model, and then evaluates positioning error according to the speed error, so as to greatly improve the reliability, stability and accuracy of the fusion positioning algorithm.

Description

Technical Field

[0001] This invention belongs to the field of multi-sensor fusion positioning technology, and particularly relates to a method, apparatus, device and readable storage medium for optimizing a fusion positioning noise model. Background Technology

[0002] In complex industrial scenarios (such as gas stations), any single positioning sensor struggles to achieve accurate and reliable navigation and positioning. Based on different measurement principles, positioning sensors such as RTK GNSS (Global Navigation Satellite System), inertial navigation devices, odometers, lidar, and vision systems each have their own advantages and disadvantages, and they are highly complementary. To address the issues of low reliability, poor continuity, and insufficient robustness in integrated navigation systems, fusion positioning technology, which integrates data from multiple sensors, has become an important research direction.

[0003] In fusion positioning systems, GNSS noise model errors directly impact the accuracy of data fusion. Most commonly used GNSS noise models are based on prior values; however, dynamically changing external environments significantly affect these models. For example, in complex gas station safety inspection scenarios, there will inevitably be differences between the true and prior values ​​of the GNSS noise model parameters, leading to non-negligible system errors and even filter divergence, resulting in inaccurate positioning of the safety inspection robot. Summary of the Invention

[0004] Based on this, and in response to the aforementioned technical problems, a method, apparatus, device, and readable storage medium for optimizing a fusion localization noise model are provided.

[0005] The technical solution adopted in this invention is as follows:

[0006] As a first aspect of the present invention, a method for optimizing a fusion localization noise model is provided, comprising:

[0007] Real-time acquisition of absolute values ​​of GNSS ground velocity measurements and the odometer's ground speed measurement value Calculate the velocity residuals of both, and then calculate the residual variance using the maximum likelihood estimation method shown in the following formula.

[0008]

[0009] Where x is the velocity residual, μ represents the expected value of x, σ represents the standard deviation, and θ represents the maximum likelihood estimate of the data sequence D. Represents the expectation of the maximum likelihood estimate;

[0010] coordinates of Substituting x into the convergent function y = f t (x), if The coordinates then belong to the approximate Gaussian convergence region. As GNSS velocity error variance if The coordinates then belong to the approximate Gaussian fitting region. The velocity residual is input into a pre-trained GRU network to obtain the GNSS velocity error variance. The process of generating the convergence function includes:

[0011] a. Establish multiple trend curves representing the variation of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement value with the true value of odometer ground velocity under different GNSS noise models. The different GNSS noise models refer to GNSS noise models with different GNSS velocity error standard deviations σ. The x-axis of the coordinate system where the trend curves are located is the true value of odometer ground velocity, and the y-axis is the variance of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement value.

[0012] b. Each trend curve eventually converges to the variance σ. 2 Select a series of threshold coordinate points (x1, y1), (x2, y2)...(x2, y2) from each trend curve. n ,y n The threshold coordinate point, which serves as the feature point for function convergence, converges to the variance σ. 2 The starting coordinates of the horizontal line;

[0013] c. Using the convergence feature points of the function, fit and generate a convergent function y = f that is positively correlated with the horizontal and vertical coordinates. t (x), y≥f t The dataset region of (x) is defined as the approximate Gaussian fitting region, where y < f t The dataset region of (x) is defined as the approximate Gaussian convergence region;

[0014] The GNSS velocity error variance The noise matrix parameters in the GNSS noise model are optimized as an optimization factor using the following formula:

[0015]

[0016] in, Represents the optimized noise matrix parameters. represents the noise matrix parameters before optimization, and β represents the trust factor, which ranges from 0 to 1.

[0017] As a second aspect of the present invention, a device for optimizing a fused localization noise model is provided, comprising:

[0018] The residual variance calculation module is used to obtain the absolute value of GNSS ground velocity measurements in real time. and the odometer's ground speed measurement value Calculate the velocity residuals of both, and then calculate the residual variance using the maximum likelihood estimation method shown in the following formula.

[0019]

[0020] Where x is the velocity residual, μ represents the expected value of x, σ represents the standard deviation, and θ represents the maximum likelihood estimate of the data sequence D. Represents the expectation of the maximum likelihood estimate;

[0021] The GNSS velocity error variance determination module is used to determine the coordinates. of Substituting x into the convergent function y = f t (x), if The coordinates then belong to the approximate Gaussian convergence region. As GNSS velocity error variance if The coordinates then belong to the approximate Gaussian fitting region. The velocity residual is input into a pre-trained GRU network to obtain the GNSS velocity error variance. The process of generating the convergence function includes:

[0022] a. Establish multiple trend curves representing the variation of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement value with the true value of odometer ground velocity under different GNSS noise models. The different GNSS noise models refer to GNSS noise models with different GNSS velocity error standard deviations σ. The x-axis of the coordinate system where the trend curves are located is the true value of odometer ground velocity, and the y-axis is the variance of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement value.

[0023] b. Each trend curve eventually converges to the variance σ. 2 Select a series of threshold coordinate points (x1, y1), (x2, y2)...(x2, y2) from each trend curve. n ,y n The threshold coordinate point, which serves as the feature point for function convergence, converges to the variance σ. 2 The starting coordinates of the horizontal line;

[0024] c. Using the convergence feature points of the function, fit and generate a convergent function y = f that is positively correlated with the horizontal and vertical coordinates.t (x), y≥f t The dataset region of (x) is defined as the approximate Gaussian fitting region, where y < f t The dataset region of (x) is defined as the approximate Gaussian convergence region;

[0025] The optimization module is used to optimize the variance of the GNSS velocity error. The noise matrix parameters in the GNSS noise model are optimized as an optimization factor using the following formula:

[0026]

[0027] in, Represents the optimized noise matrix parameters. represents the noise matrix parameters before optimization, and β represents the trust factor, which ranges from 0 to 1.

[0028] As a third aspect of the present invention, an electronic device is provided, including a storage module, the storage module including instructions loaded and executed by a processor, the instructions, when executed, causing the processor to perform a fusion localization noise model optimization method of the first aspect described above.

[0029] As a fourth aspect of the present invention, a computer-readable storage medium is provided that stores one or more programs, which, when executed by a processor, implement the fusion localization noise model optimization method of the first aspect described above.

[0030] This invention is based on an approximate Gaussian model. It uses the trend curves of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement as a function of the true value of the odometer ground velocity under different GNSS noise models to fit and generate a convergent function y = f. t (x), and based on this convergence function, define the approximate Gaussian fitting region and the approximate Gaussian convergence region. During the operation of the fusion positioning system, firstly, the absolute value of GNSS ground velocity measurement is acquired in real time. and the odometer's ground speed measurement value Calculate the velocity residuals of the two variables, and then use the maximum likelihood estimation method to calculate the residual variance. Then, when coordinates When it belongs to the approximately Gaussian convergence region, As GNSS velocity error variance When coordinates When the region falls within the approximate Gaussian fitting range, the velocity residuals are input into a pre-trained GRU network to obtain the GNSS velocity error variance. Finally, through GNSS velocity error variance By optimizing the noise matrix parameters, a real-time GNSS measurement noise model is obtained. This allows for real-time optimization of the GNSS noise model in dynamic environments. The optimized GNSS real-time measurement noise model can then be used to assess velocity errors, and consequently, to assess positioning errors. This significantly improves the reliability, stability, and accuracy of the fusion positioning algorithm. Attached Figure Description

[0031] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments:

[0032] Figure 1 A flowchart of a fusion localization noise model optimization method provided in an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram of a fusion localization noise model optimization device provided in an embodiment of the present invention;

[0034] Figure 3 A schematic diagram of an electronic device provided in an embodiment of the present invention;

[0035] Figure 4 This is a flowchart illustrating the fusion positioning process in an embodiment of the present invention;

[0036] Figure 5 This is a trend curve representing an embodiment of the present invention;

[0037] Figure 6 This is a schematic diagram of the Gaussian region according to an embodiment of the present invention. Detailed Implementation

[0038] The embodiments of the present invention will be described below with reference to the accompanying drawings. It should be noted that the embodiments described in this specification are not exhaustive and do not represent the only embodiments of the present invention. The corresponding embodiments below are only for clearly illustrating the inventive content of this patent and are not intended to limit its implementation. For those skilled in the art, different variations and modifications can be made based on the embodiments described. Any variations or modifications that fall within the technical concept and inventive content of this invention and are obvious are also within the protection scope of this invention.

[0039] Figure 4 The fusion localization process for a security inspection robot is illustrated:

[0040] S201. Evaluate the speed error of the safety inspection robot using a GNSS noise model.

[0041] S202. Evaluate the positioning error of the safety inspection robot based on the speed error.

[0042] S203. Position the safety inspection robot based on the positioning error.

[0043] Of course, the above process can also be used to perform fusion positioning of other moving targets.

[0044] It should be noted that there is no corresponding benchmark to evaluate the position of a moving target. However, the error sources of GNSS velocity error and positioning error are the same, and the velocity error can be evaluated using odometer as a benchmark. Therefore, the positioning error can be evaluated based on the velocity error.

[0045] In existing technologies, the parameters of GNSS noise models are fixed (prior values).

[0046] like Figure 1 As shown in the figure, this embodiment of the invention provides a method for optimizing a fusion positioning noise model, used to optimize the parameters of a GNSS noise model. The specific process is as follows:

[0047] S101. Real-time acquisition of absolute values ​​of GNSS ground velocity measurements. and the odometer's ground speed measurement value Calculate the velocity residuals of the two variables, and then calculate the residual variance using the maximum likelihood estimation method shown in equation (1).

[0048]

[0049] Where x is the velocity residual mentioned above, and the velocity residual refers to the absolute value of the GNSS ground velocity measurement. Ground speed measurement value of odometer The difference, μ represents the expected value of x, σ represents the standard deviation, and θ represents the maximum likelihood estimate of the data sequence D. This represents the expectation of the maximum likelihood estimate.

[0050] GNSS measurements (output values) are eastward and northward velocities. Equation (2) can be used to convert GNSS measurements into absolute values ​​of ground velocity.

[0051]

[0052] Among them, v E and v N These are the true values ​​of the GNSS eastward and northward velocities, δv. E and δv N These are the GNSS eastward and northward velocity errors, respectively. It should be noted that the measured value = true value + error.

[0053]

[0054] From equation (3), the following statistical characteristics can be obtained: GNSS velocity error δvE δv N and odometer speed error δv od All of them conform to Gaussian distribution, but their residuals are a linear superposition of Gaussian and chi-square distributions. When the true value of velocity is 0, the residuals are a linear superposition of Rayleigh and Gaussian distributions. When the true value of velocity is much larger than the error, the residuals can be approximated as a linear superposition of multiple Gaussian distributions.

[0055] S102. The coordinates (the above-mentioned odometer ground velocity measurement values, the above-mentioned residual variance), i.e., the coordinates of Substituting x into the convergent function y = f t (x), based on the above statistical properties, if This coordinate belongs to the approximately Gaussian convergence region, meaning the velocity residuals approximately conform to a Gaussian distribution. Therefore, the aforementioned residual variance... As GNSS velocity error variance if This coordinate falls within the approximate Gaussian fitting region, indicating that the velocity residual does not conform to a Gaussian distribution. Therefore, the velocity residual is input into the pre-trained GRU network to obtain the GNSS velocity error variance. Wherein, the convergent function y = f t The generation process of (x) includes:

[0056] a. Establish multiple trend curves representing the variation of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement under different GNSS noise models, as the odometer ground velocity true value changes. Different GNSS noise models refer to GNSS noise models with different GNSS velocity error standard deviations σ. The x-axis of the coordinate system containing the trend curves represents the true value of the odometer ground velocity, and the y-axis represents the variance of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement. In this embodiment, the specific process of step a is as follows:

[0057] I. GNSS velocity data simulation:

[0058] 1) Generate eastward velocity truth arrays and northward velocity truth arrays of equal length, and combine them to obtain the GNSS velocity truth matrix, the length of which is K (K = 600):

[0059] Generate a true array of eastward velocities of length 600 from 0 to 5.99 m / s, with intervals of 0.01. E ) 600×1 ];

[0060] Generate a true array of northbound velocities of length 600 from 0 to 2.995 m / s, with intervals of 0.005. N ) 600×1 ];

[0061] [(v E ) 600×1 ] and [(v N ) 600×1 The GNSS velocity truth matrix is ​​obtained by combining [(v E ) 600×1 (v N ) 600×1 ].

[0062] 2) Generate multiple sets of GNSS velocity error standard deviations. Each set of GNSS velocity error standard deviations includes the eastward velocity error standard deviation σ. E and the standard deviation of the northbound velocity error σ N :

[0063] Nine sets of velocity error standard deviations [(σ] were generated at intervals of 0.1 m / s to 0.9 m / s. E ) 9×1 (σ N ) 9×1 ].

[0064] 3) Based on the standard deviation of the GNSS velocity error for each group, generate a length C (C=1000), with a mean of 0 and variances of (σ) E ) 2 and (σ) N ) 2 Gaussian noise and We obtain G (G=9) GNSS velocity error matrices: Where i is 1, 2, 3...9.

[0065] 4) Take each row (1……m……600) of the GNSS velocity truth matrix and expand it into an expanded matrix of length C (C=1000) and 2 columns according to the same mean and variance, for a total of K expanded matrices, i.e. 600 expanded matrices. Then, take one GNSS velocity error matrix in turn and superimpose it with all the expanded matrices to obtain K GNSS velocity measurement matrixes relative to each GNSS velocity error matrix, for a total of K*G GNSS velocity measurement matrixes, as shown in equation (4):

[0066]

[0067] in, and This represents the element in the m-th row of the measurement matrix.

[0068] 5) Convert each GNSS velocity measurement matrix into an array of absolute values ​​of GNSS ground velocity measurements, resulting in a total of K*G GNSS ground velocity measurement absolute value arrays. The corresponding values ​​can then be calculated using equation (2). GNSS Earth velocity absolute value array

[0069] II. Odometer Data Simulation:

[0070] 1) The true value matrix of GNSS velocity [(v] is obtained by equation (5). E ) 600×1 (v N ) 600×1 Convert [v] to a ground velocity truth array of length K (K = 600) [(v od ) 600×1 ]:

[0071]

[0072] 2) Based on the preset standard deviation σ of the odometer error od (e.g., 0.02 m / s), with a generation length of C (C = 1000), a mean of 0, and a variance of (σ). od ) 2 Gaussian noise was used to obtain the odometer error array [δv] od ] 1000×1 ;

[0073] 3) Retrieve the ground velocity truth array [(v od ) 600×1 Each element of ][(v od ) n (1……n……600), each array is expanded into an expanded array of length C (C=1000) according to the same mean and variance. There are K extended arrays in total, totaling 600 extended arrays, and each extended array is linked to the odometer error array [δv]. od ] 1000×1 By superimposing the data, we obtain an array of K odometer-to-ground velocity measurements.

[0074] III. Array the corresponding absolute values ​​of GNSS ground velocity measurements array of odometry ground velocity measurements Subtracting (m and n are both from 1 to 600, subtracting each term) yields the velocity residual array. Then, using equation (6) to calculate the residual variance based on the velocity residual array, a residual variance matrix of G rows and K columns, i.e., 9 rows and 600 columns, is obtained. See Equation (7), where each row of the residual variance matrix corresponds to a different GNSS noise model:

[0075]

[0076]

[0077] IV. Based on the residual variance matrix Each row of elements and the ground velocity truth array [(v od ) 600×1 Generate G trend curves, see [link / reference]. Figure 5 ,exist Figure 5 In the figure, the x-axis represents the true value of the odometer's ground velocity, and the y-axis represents the variance of the velocity residual (residual noise variance) between the absolute value of the GNSS ground velocity measurement and the odometer's ground velocity measurement.

[0078] b. Each trend curve eventually converges to the variance σ. 2 Select a series of threshold coordinate points (x1, y1), (x2, y2)...(x2, y2) from each trend curve. n ,y n ) is used as the convergence feature point of the function, where the threshold coordinate point converges to the variance σ. 2 The starting coordinates of the horizontal line.

[0079] c. Using the convergence feature points of the above function, a convergent function y = f is fitted and generated that is positively correlated with the horizontal and vertical coordinates. t (x), y≥f t The dataset region of (x) is defined as the approximate Gaussian fitting region ((x,y)∈G), where y<f t The dataset region of (x) is defined as the approximate Gaussian convergence region ((x,y)∈C), see [link to relevant documentation]. Figure 6 .

[0080] The main feature of the GRU network is that it merges the forget gate and the input gate into an update gate. The update gate and the reset gate are used to predict and estimate the parameters, which greatly reduces the complexity of the model and improves the practical effect.

[0081] In this embodiment, the residual variance matrix of the above simulation data is used. and the array of odometer ground speed measurements = The input is fed into the GRU network for training. After training, the velocity residuals are input into the GRU network, and the GRU network can output the standard deviation σ of the GNSS velocity error. E (σ E =σ N Thus, the variance of GNSS velocity error is obtained. That is, the square of the standard deviation of GNSS velocity error.

[0082] S103, GNSS velocity error variance As an optimization factor, the noise matrix parameters in the GNSS noise model are optimized using equation (8):

[0083]

[0084] in, Represents the optimized noise matrix parameters. represents the noise matrix parameters before optimization, and β represents the trust factor, which ranges from 0 to 1.

[0085] By optimizing the noise matrix parameters, a real-time GNSS measurement noise model can be obtained.

[0086] As can be seen from the above, the method in this embodiment is based on an approximate Gaussian model. It uses the trend curves of the velocity residual between the absolute value of the GNSS ground velocity measurement and the odometer ground velocity measurement under different GNSS noise models to fit and generate a convergent function y = f. t (x), and based on this convergence function, define the approximate Gaussian fitting region and the approximate Gaussian convergence region. During the operation of the fusion positioning system, firstly, the absolute value of GNSS ground velocity measurement is acquired in real time. = and the odometer's ground speed measurement value Calculate the velocity residuals of the two variables, and then use the maximum likelihood estimation method to calculate the residual variance. Then, when coordinates When it belongs to the approximately Gaussian convergence region, As GNSS velocity error variance When coordinates When the region falls within the approximate Gaussian fitting range, the velocity residuals are input into a pre-trained GRU network to obtain the GNSS velocity error variance. Finally, through GNSS velocity error variance By optimizing the noise matrix parameters, a real-time GNSS measurement noise model is obtained. This allows for real-time optimization of the GNSS noise model in dynamic environments. The optimized GNSS real-time measurement noise model can then be used to assess velocity errors, and consequently, to assess positioning errors. This significantly improves the reliability, stability, and accuracy of the fusion positioning algorithm.

[0087] The following will describe in detail one or more embodiments of the fusion localization noise model optimization apparatus of the present invention. Those skilled in the art will understand that these optimization apparatuses can be configured using commercially available hardware components through the steps taught in this solution. Figure 2 This invention illustrates a fusion localization noise model optimization device provided by an embodiment of the present invention, such as... Figure 2 As shown, the optimization device includes a residual variance calculation module 11, a GNSS velocity error variance determination module 12, and an optimization module 13.

[0088] The residual variance calculation module 11 is used to obtain the absolute value of GNSS ground velocity measurements in real time. and the odometer's ground speed measurement value Calculate the velocity residuals of the two variables, and then calculate the residual variance using the maximum likelihood estimation method shown in equation (1).

[0089]

[0090] Where x is the velocity residual mentioned above, and the velocity residual refers to the absolute value of the GNSS ground velocity measurement. Ground speed measurement value of odometer The difference, μ represents the expected value of x, σ represents the standard deviation, and θ represents the maximum likelihood estimate of the data sequence D. This represents the expectation of the maximum likelihood estimate.

[0091] GNSS measurements (output values) are eastward and northward velocities. Equation (2) can be used to convert GNSS measurements into absolute values ​​of ground velocity.

[0092]

[0093] Among them, v E and v N These are the true values ​​of the GNSS eastward and northward velocities, δv. E and δv N These are the GNSS eastward and northward velocity errors, respectively. It should be noted that the measured value = true value + error.

[0094]

[0095] From equation (3), the following statistical characteristics can be obtained: GNSS velocity error δv E δv N and odometer speed error δv od All of them conform to Gaussian distribution, but their residuals are a linear superposition of Gaussian and chi-square distributions. When the true value of velocity is 0, the residuals are a linear superposition of Rayleigh and Gaussian distributions. When the true value of velocity is much larger than the error, the residuals can be approximated as a linear superposition of multiple Gaussian distributions.

[0096] S102. The coordinates (the above-mentioned odometer ground velocity measurement values, the above-mentioned residual variance), i.e., the coordinates of Substituting x into the convergent function y = f t (x), based on the above statistical properties, if This coordinate belongs to the approximately Gaussian convergence region, meaning the velocity residuals approximately conform to a Gaussian distribution. Therefore, the aforementioned residual variance... As GNSS velocity error variance if This coordinate falls within the approximate Gaussian fitting region, indicating that the velocity residual does not conform to a Gaussian distribution. Therefore, the velocity residual is input into the pre-trained GRU network to obtain the GNSS velocity error variance. Wherein, the convergent function y = f t The generation process of (x) includes:

[0097] a. Establish multiple trend curves representing the variation of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement under different GNSS noise models, as the odometer ground velocity true value changes. Different GNSS noise models refer to GNSS noise models with different GNSS velocity error standard deviations σ. The x-axis of the coordinate system containing the trend curves represents the true value of the odometer ground velocity, and the y-axis represents the variance of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement. In this embodiment, the specific process of step a is as follows:

[0098] I. GNSS velocity data simulation:

[0099] 1) Generate eastward velocity truth arrays and northward velocity truth arrays of equal length, and combine them to obtain the GNSS velocity truth matrix, the length of which is K (K = 600):

[0100] Generate a true array of eastward velocities of length 600 from 0 to 5.99 m / s, with intervals of 0.01. E ) 600×1 ];

[0101] Generate a true array of northbound velocities of length 600 from 0 to 2.995 m / s, with intervals of 0.005. N ) 600×1 ];

[0102] [(v E ) 600×1 ] and [(v N ) 600×1 The GNSS velocity truth matrix is ​​obtained by combining [(v E ) 600×1 (v N ) 600×1 ].

[0103] 2) Generate multiple sets of GNSS velocity error standard deviations. Each set of GNSS velocity error standard deviations includes the eastward velocity error standard deviation σ. E and the standard deviation of the northbound velocity error σ N :

[0104] Nine sets of velocity error standard deviations [(σ] were generated at intervals of 0.1 m / s to 0.9 m / s. E ) 9×1 (σN ) 9×1 ].

[0105] 3) Based on the standard deviation of the GNSS velocity error for each group, generate a length C (C=1000), with a mean of 0 and variances of (σ) E ) 2 and (σ) N ) 2 Gaussian noise and We obtain G (G=9) GNSS velocity error matrices: Where i is 1, 2, 3...9.

[0106] 4) Take each row (1……m……600) of the GNSS velocity truth matrix and expand it into an expanded matrix of length C (C=1000) and 2 columns according to the same mean and variance, for a total of K expanded matrices, i.e. 600 expanded matrices. Then, take one GNSS velocity error matrix in turn and superimpose it with all the expanded matrices to obtain K GNSS velocity measurement matrixes relative to each GNSS velocity error matrix, for a total of K*G GNSS velocity measurement matrixes, as shown in equation (4):

[0107]

[0108] in, and This represents the element in the m-th row of the measurement matrix.

[0109] 5) Convert each GNSS velocity measurement matrix into an array of absolute values ​​of GNSS ground velocity measurements, resulting in a total of K*G GNSS ground velocity measurement absolute value arrays. The corresponding values ​​can then be calculated using equation (2). GNSS Earth velocity absolute value array

[0110] II. Odometer Data Simulation:

[0111] 1) The true value matrix of GNSS velocity [(v] is obtained by equation (5). E ) 600×1 (v N ) 600×1 Convert [v] to a ground velocity truth array of length K (K = 600) [(v od ) 600×1 ]:

[0112]

[0113] 2) Based on the preset standard deviation σ of the odometer error od(e.g., 0.02 m / s), with a generation length of C (C = 1000), a mean of 0, and a variance of (σ). od ) 2 Gaussian noise was used to obtain the odometer error array [δv] od ] 1000×1 ;

[0114] 3) Retrieve the ground velocity truth array [(v od ) 600×1 Each element of ][(v od ) n (1……n……600), each array is expanded into an expanded array of length C (C=1000) according to the same mean and variance. There are K extended arrays in total, totaling 600 extended arrays, and each extended array is linked to the odometer error array [δv]. od ] 1000×1 By superimposing the data, we obtain an array of K odometer-to-ground velocity measurements.

[0115] III. Array the corresponding absolute values ​​of GNSS ground velocity measurements array of odometry ground velocity measurements Subtracting (m and n are both from 1 to 600, subtracting each term) yields the velocity residual array. Then, using equation (6) to calculate the residual variance based on the velocity residual array, a residual variance matrix of G rows and K columns, i.e., 9 rows and 600 columns, is obtained. See Equation (7), where each row of the residual variance matrix corresponds to a different GNSS noise model:

[0116]

[0117]

[0118] IV. Based on the residual variance matrix Each row of elements and the ground velocity truth array [(v od ) 600×1 Generate G trend curves, see [link / reference]. Figure 5 ,exist Figure 5 In the figure, the x-axis represents the true value of the odometer's ground velocity, and the y-axis represents the variance of the velocity residual (residual noise variance) between the absolute value of the GNSS ground velocity measurement and the odometer's ground velocity measurement.

[0119] b. Each trend curve eventually converges to the variance σ. 2 Select a series of threshold coordinate points (x1, y1), (x2, y2)...(x2, y2) from each trend curve. n ,y n) is used as the convergence feature point of the function, where the threshold coordinate point converges to the variance σ. 2 The starting coordinates of the horizontal line.

[0120] c. Using the convergence feature points of the above function, a convergent function y = f is fitted and generated that is positively correlated with the horizontal and vertical coordinates. t (x), y≥f t The dataset region of (x) is defined as the approximate Gaussian fitting region ((x,y)∈G), where y<f t The dataset region of (x) is defined as the approximate Gaussian convergence region ((x,y)∈C), see [link to relevant documentation]. Figure 6 .

[0121] The main feature of the GRU network is that it merges the forget gate and the input gate into an update gate. The update gate and the reset gate are used to predict and estimate the parameters, which greatly reduces the complexity of the model and improves the practical effect.

[0122] In this embodiment, the residual variance matrix of the above simulation data is used. and the array of odometer ground speed measurements = The input is fed into the GRU network for training. After training, the velocity residuals are input into the GRU network, and the GRU network can output the standard deviation σ of the GNSS velocity error. E (σ E =σ N Thus, the variance of GNSS velocity error is obtained. That is, the square of the standard deviation of GNSS velocity error.

[0123] Optimization module 13 is used to optimize the variance of GNSS velocity error. As an optimization factor, the noise matrix parameters in the GNSS noise model are optimized using equation (8):

[0124]

[0125] in, Represents the optimized noise matrix parameters. represents the noise matrix parameters before optimization, and β represents the trust factor, which ranges from 0 to 1.

[0126] By optimizing the noise matrix parameters, a real-time GNSS measurement noise model can be obtained.

[0127] In summary, the fusion positioning noise model optimization device provided in the above embodiments can execute the fusion positioning noise model optimization method provided in the foregoing embodiments.

[0128] Similar to the above concept, the above Figure 2 The structure of the underground intelligent broadcasting device shown can be implemented as an electronic device. Figure 3 A schematic block diagram of the structure of an electronic device provided by an embodiment of the present invention is shown.

[0129] For example, the electronic device includes a storage module 21 and a processor 22. The storage module 21 includes instructions loaded and executed by the processor 22, which, when executed, cause the processor 22 to perform the steps described in the section on a fusion localization noise model optimization method described above in this specification, according to various exemplary embodiments of the present invention.

[0130] It should be understood that processor 22 can be a Central Processing Unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, the general-purpose processor can be a microprocessor or any conventional processor.

[0131] This invention also provides a computer-readable storage medium that stores one or more programs, which, when executed by a processor, implement the steps described in the above-described U-shaped intelligent broadcasting method section according to various exemplary embodiments of the invention.

[0132] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer-readable storage media (or non-transitory media) and communication media (or transient media).

[0133] As is known to those skilled in the art, the term computer-readable storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer-readable storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0134] For example, the computer-readable storage medium may be an internal storage module of the electronic device described in the foregoing embodiments, such as a hard drive or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard drive, SmartMediaCard (SMC), SecureDigital (SD) card, FlashCard, etc., provided on the electronic device.

[0135] The electronic devices and computer-readable storage media provided in the foregoing embodiments, based on an approximate Gaussian model, use the trend curves of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement under different GNSS noise models to fit and generate a convergent function y = f t (x), and based on this convergence function, define the approximate Gaussian fitting region and the approximate Gaussian convergence region. During the operation of the fusion positioning system, firstly, the absolute value of GNSS ground velocity measurement is acquired in real time. and the odometer's ground speed measurement value Calculate the velocity residuals of the two variables, and then use the maximum likelihood estimation method to calculate the residual variance. Then, when coordinates When it belongs to the approximately Gaussian convergence region, As GNSS velocity error variance When coordinates When the region falls within the approximate Gaussian fitting range, the velocity residuals are input into a pre-trained GRU network to obtain the GNSS velocity error variance. Finally, through GNSS velocity error variance By optimizing the noise matrix parameters, a real-time GNSS measurement noise model is obtained. This allows for real-time optimization of the GNSS noise model in dynamic environments. The optimized GNSS real-time measurement noise model can then be used to assess velocity errors, and consequently, to assess positioning errors. This significantly improves the reliability, stability, and accuracy of the fusion positioning algorithm.

[0136] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for optimizing a fusion localization noise model, characterized in that, include: Real-time acquisition of absolute values ​​of GNSS ground velocity measurements and the odometer's ground speed measurement value Calculate the velocity residuals of both, and then calculate the residual variance using the maximum likelihood estimation method shown in the following formula. Where x is the velocity residual, μ represents the expected value of x, σ represents the standard deviation, and θ represents the maximum likelihood estimate of the data sequence D. Represents the expectation of the maximum likelihood estimate; coordinates of Substituting x into the convergent function y = f t (x), if The coordinates then belong to the approximate Gaussian convergence region. As GNSS velocity error variance if The coordinates then belong to the approximate Gaussian fitting region. The velocity residual is input into a pre-trained GRU network to obtain the GNSS velocity error variance. The process of generating the convergence function includes: a. Establish multiple trend curves representing the variation of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement value with the true value of odometer ground velocity under different GNSS noise models. The different GNSS noise models refer to GNSS noise models with different GNSS velocity error standard deviations σ. The x-axis of the coordinate system where the trend curves are located is the true value of odometer ground velocity, and the y-axis is the variance of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement value. b. Each trend curve eventually converges to the variance σ. 2 Select a series of threshold coordinate points (x1, y1), (x2, y2)...(x2, y2) from each trend curve. n ,y n The threshold coordinate point, which serves as the feature point for function convergence, converges to the variance σ. 2 The starting coordinates of the horizontal line; c. Using the convergence feature points of the function, fit and generate a convergent function y = f that is positively correlated with the horizontal and vertical coordinates. t (x), y≥f t The dataset region of (x) is defined as the approximate Gaussian fitting region, where y < f t The dataset region of (x) is defined as the approximate Gaussian convergence region; The GNSS velocity error variance The noise matrix parameters in the GNSS noise model are optimized as an optimization factor using the following formula: in, Represents the optimized noise matrix parameters. represents the noise matrix parameters before optimization, and β represents the trust factor, which ranges from 0 to 1.

2. The method for optimizing a fusion localization noise model according to claim 1, characterized in that, Step a further includes: GNSS velocity data simulation: 1) Generate eastward velocity truth arrays and northward velocity truth arrays of equal length, and combine the two to obtain the GNSS velocity truth matrix, the length of which is K; 2) Generate multiple sets of GNSS velocity error standard deviations. Each set of GNSS velocity error standard deviations includes the eastward velocity error standard deviation σ. E and the standard deviation of the northbound velocity error σ N ; 3) Based on the standard deviation of the GNSS velocity error for each group, generate a generator with length C, mean 0, and variances of (σ... E ) 2 and (σ) N ) 2 Gaussian noise is used to obtain G GNSS velocity error matrices; 4) Take each row of the GNSS velocity truth matrix and expand it into an expanded matrix of length C and 2 columns, for a total of K expanded matrices. Then, take a GNSS velocity error matrix and superimpose it with all the expanded matrices to obtain K GNSS velocity measurement matrices relative to each GNSS velocity error matrix, for a total of K*G GNSS velocity measurement matrices. 5) Convert each GNSS velocity measurement matrix into an array of absolute values ​​of GNSS ground velocity measurements, resulting in a total of K*G arrays of absolute values ​​of GNSS ground velocity measurements; Odometer data simulation: 1) Convert the GNSS velocity truth matrix into a ground velocity truth array of length K; 2) Based on the preset standard deviation σ of the odometer error od Generate a length of C, a mean of 0, and a variance of (σ). od ) 2 Gaussian noise was used to obtain the odometer error array; 3) Take each element of the ground velocity true value array and expand it into an expanded array of length C, for a total of K expanded arrays. Then, superimpose each expanded array with the odometer error array to obtain K odometer ground velocity measurement value arrays. Subtract the corresponding GNSS ground velocity absolute value array from the odometer ground velocity value array to obtain the velocity residual array, and calculate the residual variance based on the velocity residual array to obtain a residual variance matrix of G rows and K columns. Each row of the residual variance matrix corresponds to a different GNSS noise model. Based on each row element of the residual variance matrix and the ground velocity truth array, G trend curves are generated.

3. The method for optimizing a fusion localization noise model according to claim 2, characterized in that, The residual variance matrix and the odometry ground velocity measurement array are input into the GRU network for training.

4. A device for optimizing a fused localization noise model, characterized in that, include: The residual variance calculation module is used to obtain the absolute value of GNSS ground velocity measurements in real time. and the odometer's ground speed measurement value Calculate the velocity residuals of both, and then calculate the residual variance using the maximum likelihood estimation method shown in the following formula. Where x is the velocity residual, μ represents the expected value of x, σ represents the standard deviation, and θ represents the maximum likelihood estimate of the data sequence D. Represents the expectation of the maximum likelihood estimate; The GNSS velocity error variance determination module is used to determine the coordinates. of Substituting x into the convergent function y = f t (x), if The coordinates then belong to the approximate Gaussian convergence region. As GNSS velocity error variance if The coordinates then belong to the approximate Gaussian fitting region. The velocity residual is input into a pre-trained GRU network to obtain the GNSS velocity error variance. The process of generating the convergence function includes: a. Establish multiple trend curves representing the variation of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement value with the true value of odometer ground velocity under different GNSS noise models. The different GNSS noise models refer to GNSS noise models with different GNSS velocity error standard deviations σ. The x-axis of the coordinate system where the trend curves are located is the true value of odometer ground velocity, and the y-axis is the variance of the velocity residual between the absolute value of GNSS ground velocity measurement and the odometer ground velocity measurement value. b. Each trend curve eventually converges to the variance σ. 2 Select a series of threshold coordinate points (x1, y1), (x2, y2)...(x2, y2) from each trend curve. n ,y n The threshold coordinate point, which serves as the feature point for function convergence, converges to the variance σ. 2 The starting coordinates of the horizontal line; c. Using the convergence feature points of the function, fit and generate a convergent function y = f that is positively correlated with the horizontal and vertical coordinates. t (x), y≥f t The dataset region of (x) is defined as the approximate Gaussian fitting region, where y < f t The dataset region of (x) is defined as the approximate Gaussian convergence region; The optimization module is used to optimize the variance of the GNSS velocity error. The noise matrix parameters in the GNSS noise model are optimized as an optimization factor using the following formula: in, Represents the optimized noise matrix parameters. represents the noise matrix parameters before optimization, and β represents the trust factor, which ranges from 0 to 1.

5. The fusion localization noise model optimization device according to claim 4, characterized in that, Step a further includes: GNSS velocity data simulation: 1) Generate eastward velocity truth arrays and northward velocity truth arrays of equal length, and combine the two to obtain the GNSS velocity truth matrix, the length of which is K; 2) Generate multiple sets of GNSS velocity error standard deviations. Each set of GNSS velocity error standard deviations includes the eastward velocity error standard deviation σ. E and the standard deviation of the northbound velocity error σ N ; 3) Based on the standard deviation of the GNSS velocity error for each group, generate a generator with length C, mean 0, and variances of (σ... E ) 2 and (σ) N ) 2 Gaussian noise is used to obtain G GNSS velocity error matrices; 4) Take each row of the GNSS velocity truth matrix and expand it into an expanded matrix of length C and 2 columns, for a total of K expanded matrices. Then, take a GNSS velocity error matrix and superimpose it with all the expanded matrices to obtain K GNSS velocity measurement matrices relative to each GNSS velocity error matrix, for a total of K*G GNSS velocity measurement matrices. 5) Convert each GNSS velocity measurement matrix into an array of absolute values ​​of GNSS ground velocity measurements, resulting in a total of K*G arrays of absolute values ​​of GNSS ground velocity measurements; Odometer data simulation: 1) Convert the GNSS velocity truth matrix into a ground velocity truth array of length K; 2) Based on the preset standard deviation σ of the odometer error od Generate a length of C, a mean of 0, and a variance of (σ). od ) 2 Gaussian noise was used to obtain the odometer error array; 3) Take each element of the ground velocity true value array and expand it into an expanded array of length C, for a total of K expanded arrays. Then, superimpose each expanded array with the odometer error array to obtain K odometer ground velocity measurement value arrays. Subtract the corresponding GNSS ground velocity absolute value array from the odometer ground velocity value array to obtain the velocity residual array, and calculate the residual variance based on the velocity residual array to obtain a residual variance matrix of G rows and K columns. Each row of the residual variance matrix corresponds to a different GNSS noise model. Based on each row element of the residual variance matrix and the ground velocity truth array, G trend curves are generated.

6. The fusion localization noise model optimization device according to claim 5, characterized in that, The residual variance matrix and the odometry ground velocity measurement array are input into the GRU network for training.

7. An electronic device, characterized in that, The system includes a storage module comprising instructions loaded and executed by a processor, which, when executed, cause the processor to perform a fusion localization noise model optimization method according to any one of claims 1-3.

8. A computer-readable storage medium storing one or more programs, characterized in that, When the one or more programs are executed by the processor, they implement the fusion localization noise model optimization method according to any one of claims 1-3.

Images